Collaboration and social networking are increasingly important for academics, yet identifying relevant collaborators requires remarkable effort. While there are various networking services optimized for seeking similarities between the users, the scholarly motive of producing new knowledge calls for assistance in identifying people with complementary qualities. However, there is little empirical understanding of how academics perceive relevance, complementarity, and diversity of individuals in their profession and how these concepts can be optimally embedded in social matching systems. This paper aims to support the development of diversity-enhancing people recommender systems by exploring senior researchers’ perceptions of recommended other scholars at different levels on a similar–different continuum. To conduct the study, we built a recommender system based on topic modeling of scholars’ publications in the DBLP computer science bibliography. A study of 18 senior researchers comprised a controlled experiment and semi-structured interviewing, focusing on their subjective perceptions regarding relevance, similarity, and familiarity of the given recommendations, as well as participants’ readiness to interact with the recommended people. The study implies that the homophily bias (behavioral tendency to select similar others) is strong despite the recognized need for complementarity. While the experiment indicated consistent and significant differences between the perceived relevance of most similar vs. other levels, the interview results imply that the evaluation of the relevance of people recommendations is complex and multifaceted. Despite the inherent bias in selection, the participants could identify highly interesting collaboration opportunities on all levels of similarity.
Scholars’ Perceptions of Relevance in Bibliography-Based People Recommender System
1. Scholars’ Perceptions of Relevance
in Bibliography-based People
Recommender System
Ekaterina Olshannikova, Thomas Olsson, Jukka Huhtamäki, Peng Yao
Tampere University, Tampere, Finland
ekaterina.olshannikova@tuni.fi
thomas.olsson@tuni.fi
jukka.huhtamaki@tuni.fi
peng.yao@tuni.fi
ECSCW 2019, Salzburg, Austria, 8 – 12 June 2019
2. Contents
1. Motivation and Research Gap
2. Experiment Design
3. Findings
4. Discussion and Conclusions
5. FutureWork
4. Academic Partnering
ACCESS TO EXPERTISE
ACCESS TO FUNDING
IDEA CREATION
KNOWLEDGE EXHANGE
1. Motivation and Research Gap 4
5. Social Matching Systems
• People Recommender Systems
• Opportunistic Matching Applications
• Expert Finding System
• Event-based Mobile Applications
• Bibliography-based Systems for Researchers
1. Motivation and Research Gap 5
7. Research Gap
• Lack of research and design for diversifying people
recommendations
• Little focus on user perceptions in evaluation of people
recommendations
1. Motivation and Research Gap 7
8. Goal and Objectives
Goal: Contribute to the development of matching methods that not
only identify maximal similarity but help to explore the
difference dimension
Objectives:
• Design bibliography-based people recommender system for
researchers
• Evaluate perceived relevance of recommendations
• Observe the homophily bias amongst scholars
1. Motivation and Research Gap 8
9. Research Questions and Hypothesis
RQ1: What level of measured similarity of publication history is
preferred in recommendations of potential collaborators?
RQ2: What specific needs and expectations scholars have in
regard to seeking professional collaboration?
H: The more similar recommendation is (i.e., the similarity of
publishing history), the more relevant and similar the
recommendation would be perceived.
1. Motivation and Research Gap 9
11. 1. Data source and Input
Dataset (DBLP)
(Author, Title, Year,
Conference)
Record 1
Record 2
Record n
…
INPUT
Participant Name
Participant 1
Participant 20
Conference Name
…
Conference 1
Conference n
…
…
…
Conference 1
Conference n
Low distance recom. (R1, R2, R3)
Medium distance recom. (R4, R5, R6)
High distance recom. (R7, R8, R9)
8.Output Recommendations
7. Co-Authors Network
Filter out co-authors
2. Extract target conference(s)
records for Participants
Conference 1
(Author (s), Title)
Record 1
Record 2
Record m
…
Conference n
(Author (s), Title)
… Record 1
Record 2
Record m
…
3. Clean and aggregate each authour’s records
…
Conference 1
(Author, Content)
Record 1
Record m
…
Conference n
(Author, Content)
Record 1
Record m
…
4. Run TF-IDF
TF-IDF
Feature Vector
Record 1
Record k
…
5. Sorting Cosine Distances for Participants
…
Distances for
Participant 1
Distance 1
Distance i
…
Distances for
Participant n
Distance 1
Distance i
…
6. OTSU Filter to identify
recommendations
of low, medium and high distances
0.0
0 2000 4000 6000 8000 10000
0.2
0.4
0.6
0.8
1.0
End of low
distance
Mid Point
Start of
high distance
System Design
2. Experiment Design 11
12. User Interface
Research topics
Recommendation 1 Full name
Social Recommendations for The Participant 1
Co-author 1 Co-author 2 Co-author 3 Co-author 4 Co-author 5 Co-author 6
Topic 1
Topic 6
Topic 2
Topic 7
Topic 3
Topic 8
Topic 4 Topic 5
Co-authors 8
Recent publications
Authors. Publication 1 title. Year, Venue
Authors. Publication 2 title. Year, Venue
Authors. Publication 3 title. Year, Venue
8
2. Experiment Design 12
13. Participants
7 SENIOR RESEARCHERS
6 POSTDOCTORAL RESEARCHERS
4 FULL PROFESSOR
1 ASSOCIATE PROFESSOR
13 MALE AND 5 FEMALE
AGES FROM 32 TO 66, MDN 42
EXPERIENCE FROM 10 TO 46, MDN 19
2. Experiment Design 13
14. Participants’ Expertise
• HCI (10)
• gaze technologies and interactions (7)
• wireless technologies (6)
• interaction design and techniques (6)
• interfaces and information systems (5)
• usability/user experience and user-centered design (5)
• telecommunications and networking (5)
• virtual reality (3)
• wellness/health technologies (3)
2. Experiment Design 14
15. Evaluation Criteria
• Perceived relevance of the recommendation
• Perceived similarity
• Levels of perceived familiarity
• Expected willingness to interact with a recommended person
2. Experiment Design 15
21. Needs and Important Factors in Research
Collaboration
FUNDING (12 answers)
NEEDS FACTORS
KNOWLEDGE SHARING (12 answers)
CONFLICT OF INTEREST (5 answers)
RESEARCH MOBILITY (5 answers)
AFFILIATION (14 answers)
QUALITY OF RESEARCH (8 answers)
SENIORITY (7 answers)
PERSONAL CHEMISTRY (7 answers)
3. Findings 21
23. Criteria of Relevance
• Similarity of background, attitudes, values, beliefs, goals and
intentions (e.g., research aims)
• Complementarity of professional roles, skills, knowledge, and
social capital
• Compatibility for direct cooperation in terms of being mentally,
socially, morally or emotionally close to each other
• Approachability/logistics – the availability of a person for direct
or indirect interaction in terms of physical proximity as well as
social and organizational distance
4. Discussion and Conclusions 23
24. Limitations
• Limited content – only publications titles are available for
analysis
• DBLP data set provides limited understanding to the social ties
between researchers
• The evaluation was largely based on participants’ first
impression
• Participants’ were lacking a timely need for collaboration
4. Discussion and Conclusions 24
25. Conclusions
• We provided empirical findings on subjective perceptions of
people recommendation relevance
• Presented academics' needs in collaboration and factors that
might affect decision-making in choosing partners
• The homophily bias is evident also in scholars' intuitive
assessments of relevance and willingness to interact
• There is a mismatch between people's intuitive choices and
the deliberate intentions in decision-making on potential
collaborators
• We operationalized measures for subjective opinions on
people recommendations
4. Discussion and Conclusions 25
26. 5. Future Work
Exploring Structural Positions onTwitter Networks for Followee
Recommendations:
• Dormant tie – someone that the target user is already
but has never mentioned;
• Mentions-of-Mentions -- a friend-of-a-friend type of a
connection in mention network;
• A user identified to belong to the same community cluster
Future work 26
In scholarly work, collaboration has become a normative form of knowledge production. In academic research, collaboration takes place on a dyadic level between individuals, amongst research teams, as well as within international consortia. However, identifying new suitable candidates for academic collaboration requires high investment in social networking, and the disciplinary structures can prevent unexpected combinations of individuals.
Following this trend of the increasing importance of collaboration, supporting social networking and encouraging new social encounters have become central design goals in the HCI & CSCW communities.
However, the majority of professional matching systems tend to utilize a similarity-maximizing approach, providing recommendations of like-minded others with similar interests. Homogeneity produces both positive and negative effects on interpersonal communication, community formation, and knowledge work – homophily not only unifies, it also divides a network. Collaboration within a group of people with shared interests can contribute to a safe and trustworthy work environment. It has been found that researching and cooperating with diverse individuals is essential in tasks that aim to create new knowledge.
While research on diversifying item recommendations is gaining interest, few attempts have been made to match people based on diversity. The evaluation of people recommender systems and matching applications is geared toward the assessment of algorithm effectiveness, with little focus on user perceptions. Although there is well-established research on user-centered evaluation of content recommender systems, the choice of potential collaborators is significantly different and, therefore, requires contextually operationalized evaluation metrics.
To enable the gathering of such data about recommendations’ perceptions, we developed a simple DBLP-based people recommender system that provides the user with recommendations of other scholars from three different levels of similarity regarding their publication history—low, moderate and high.
With a user study that combines a controlled experiment and semi-structured interviewing, we address the following research questions: (RQ1) What level of measured similarity of publication history is preferred in recommendations of potential collaborators? (RQ2) What specific needs and expectations scholars have in regard to seeking professional collaboration?
The people recommender system runs on the subset of the parsed almost 6 mln records and depends on the input of publication venues of a given user (participant of the study). For the topic modeling we apply TF-IDF. For identifying similarity between authors -- we compute the cosine distances. Accordingly, the closest, or the most similar author to a participant will have the smallest cosine distance. Otsu filter is applied to trace three groups groups of similarity – low (very similar), medium (somewhat similar), high (very low similarity). Three recommendations that have no co-authorship with the given participant and published in the same venues are picked from each distance group as the final output, thus delivering nine recommendations in total.
A single-page web application was deployed in Firebase development platform. Personalized recommendations visualized in the form of a carousel-based list. The UI visualizes all information about authors, which DBLP data set allows to extract. The publication list represents only works from conferences where both the recommended person and the participant of the experiment have published in. Accordingly, the list of co-authors is taken from those publications only. The topics were generated through bigram analysis on the corpus profiles of each recommended person.
We recruited 18 English-speaking senior researchers who work at two university campuses in Tampere, Finland. Following the assumption that senior researchers often have more needs for finding collaborators, we limited our scope to postdoctoral researchers, professors, or otherwise senior academic positions. Overall, we had 13 male and five female participants, all based in either of the two universities in the same city. The ages vary from 32 to 66 (Median: 42, Mean: 45).
Before starting the numerical evaluation, participants were given time to explore all the recommendations and get a general overview of the alternatives. The evaluation was constructed according to four variables. The variables were operationalized based on the authors' personal experiences and qualitative research insights on academic collaboration and user experience evaluation. The interview questions were designed to obtain participants' rationale behind the scoring of recommendations as well as to reveal needs and factors that affect decision-making in academic networking practices.
Although the participants were unaware of the three different similarity distances, in their feedback they distinguish between different degrees of perceived relevance by using phrases like `very/most relevant,' `somewhat relevant/not exact match' and `irrelevant/totally irrelevant.’ Decision-making on perceived relevance was found to be influenced by the participants' estimation about how topics of recommended people match with their own -- whether they are similar, very different or complementary.
Considering our first research question about which level of system-defined similarity is preferred in participants' evaluations, the findings demonstrate the highest ratings for most similar people. Methodologically, the subjective evaluations of different similarity levels seem to consolidate the system design. So subjective perception match well with objectively defined similarity levels. So even the relatively simple analytics procedure with scarce data seemed to work sufficiently, and the publication data represented the participants' topics accurately enough.
Familiarity variable was evaluated from three perspectives. In the interview, participants addressed that already known people are less exciting recommendations (quotes 6 and 7). Even though we intentionally filtered out all the co-authors, the bibliographic data prevents understanding of the actual social relationships between researchers. Thus, in some cases (9 recommendations out of 54), the system recommended people from very close social circles.
Evaluation of the willingness to interact with recommended people reflects six predefined scenarios of face-to-face interaction or follow-up collaboration at the context of a conference. In general, participants have diversified opinions regarding the estimation of any interactions and follow-up activities. The majority of participants consider such social activities to happen naturally at conferences and do not necessarily require high investments of time in collaboration. For some, it was hard to envision willingness to interact with unfamiliar people based on papers titles and topics of the research. Some emphasize that even in a real context of visiting the conference, it might be challenging to find relevant people in a crowd and contextualized use of such recommender systems might simplify the process and encourage social interaction with unfamiliar people.
Regarding the second research question about academics' needs and expectations in professional collaboration, the results demonstrate that the optimal area on the similarity-difference continuum highly depends on the type and context of collaboration. For instance, crucial factors in direct cooperation, such as personal compatibility and similarity of attitudes and beliefs, are not as emphasized in short-term and indirect professional interactions (e.g., consultancy type of cooperation) as in long-term collaboration. Furthermore, the nature of the collaboration task might influence the perceived relevance of potential candidates, for example, regarding the complementarity of professional roles, skills, and knowledge.
Following the participants' rationale about important factors in collaboration, we propose that the diversity of recommendations in professional social matching could be enhanced through several dimensions or criteria of relevance: Previous research addressed that similarity of such qualities can raise cohesion or so-called `affinity' in interpersonal relationships, and can even reduce adverse effects of individual dissimilarities in collaborative work; identifying beneficial intersections between individuals' qualities. Collaboration is a process of knowledge production, in which compatibility of such qualities can establish trustful, joyful and personally valued cooperation.
Dorm as a recommendation mechanism, this could re-introduce users with a possibly ignored or forgotten tie; MoM in contrast to typical followership networks, the mention network is based on more explicit interactions; Com -- this could introduce new people in a programmatically identified network cluster, however not necessarily having explicit ties or interactions